agora inbox for pgsql-hackers@postgresql.orghelp / color / mirror / Atom feed
[PATCH v3] Fix DROP PROPERTY GRAPH "unsupported object class" error 604+ messages / 2 participants [nested] [flat]
* [PATCH v3] Fix DROP PROPERTY GRAPH "unsupported object class" error @ 2026-04-23 09:27 Bertrand Drouvot <bertranddrouvot.pg@gmail.com> 0 siblings, 0 replies; 604+ messages in thread From: Bertrand Drouvot @ 2026-04-23 09:27 UTC (permalink / raw) getObjectTypeDescription() and getObjectIdentityParts() are missing switch cases for PropgraphElementLabelRelationId and PropgraphLabelPropertyRelationId, causing them to hit the default case and error out with "unsupported object class". During DROP PROPERTY GRAPH, this manifests when an event trigger is active, because pg_event_trigger_ddl_commands() calls these functions. The same code paths are also reachable via pg_identify_object() and pg_identify_object_as_address(). This commit adds the missing cases. Test coverage is added in object_address.sql for these two catalog types and in create_property_graph.sql covering all property graph object types. A test in event_trigger.sql verifies that DROP PROPERTY GRAPH works correctly when an event trigger is active. Author: Bertrand Drouvot <bertranddrouvot.pg@gmail.com> Reviewed-by: Michael Paquier <michael@paquier.xyz> Reviewed-by: Ashutosh Bapat <ashutosh.bapat.oss@gmail.com> Discussion: https://postgr.es/m/aej1DkLwhyZWmtxJ@bdtpg --- src/backend/catalog/objectaddress.c | 93 +++++++++++++++++++ .../expected/create_property_graph.out | 78 ++++++++++++++++ src/test/regress/expected/event_trigger.out | 17 ++++ src/test/regress/expected/object_address.out | 10 +- .../regress/sql/create_property_graph.sql | 23 +++++ src/test/regress/sql/event_trigger.sql | 13 +++ src/test/regress/sql/object_address.sql | 6 +- 7 files changed, 238 insertions(+), 2 deletions(-) 14.7% src/backend/catalog/ 75.5% src/test/regress/expected/ 9.7% src/test/regress/sql/ diff --git a/src/backend/catalog/objectaddress.c b/src/backend/catalog/objectaddress.c index c1862809577..e2d6d8f71f6 100644 --- a/src/backend/catalog/objectaddress.c +++ b/src/backend/catalog/objectaddress.c @@ -4901,10 +4901,18 @@ getObjectTypeDescription(const ObjectAddress *object, bool missing_ok) appendStringInfoString(&buffer, "policy"); break; + case PropgraphElementLabelRelationId: + appendStringInfoString(&buffer, "property graph element label"); + break; + case PropgraphElementRelationId: appendStringInfoString(&buffer, "property graph element"); break; + case PropgraphLabelPropertyRelationId: + appendStringInfoString(&buffer, "property graph label property"); + break; + case PropgraphLabelRelationId: appendStringInfoString(&buffer, "property graph label"); break; @@ -6161,6 +6169,49 @@ getObjectIdentityParts(const ObjectAddress *object, break; } + case PropgraphElementLabelRelationId: + { + Relation ellabelDesc; + ScanKeyData skey[1]; + SysScanDesc ellabelscan; + HeapTuple tup; + Form_pg_propgraph_element_label pgelform; + ObjectAddress oa; + + ellabelDesc = table_open(PropgraphElementLabelRelationId, AccessShareLock); + ScanKeyInit(&skey[0], + Anum_pg_propgraph_element_label_oid, + BTEqualStrategyNumber, F_OIDEQ, + ObjectIdGetDatum(object->objectId)); + + ellabelscan = systable_beginscan(ellabelDesc, + PropgraphElementLabelObjectIndexId, + true, NULL, 1, skey); + + tup = systable_getnext(ellabelscan); + if (!HeapTupleIsValid(tup)) + { + if (!missing_ok) + elog(ERROR, "could not find tuple for element label %u", + object->objectId); + + systable_endscan(ellabelscan); + table_close(ellabelDesc, AccessShareLock); + break; + } + + pgelform = (Form_pg_propgraph_element_label) GETSTRUCT(tup); + + ObjectAddressSet(oa, PropgraphElementRelationId, pgelform->pgelelid); + + appendStringInfoString(&buffer, getObjectIdentityParts(&oa, objname, + objargs, false)); + + systable_endscan(ellabelscan); + table_close(ellabelDesc, AccessShareLock); + break; + } + case PropgraphElementRelationId: { HeapTuple tup; @@ -6184,6 +6235,48 @@ getObjectIdentityParts(const ObjectAddress *object, break; } + case PropgraphLabelPropertyRelationId: + { + Relation lblpropDesc; + ScanKeyData skey[1]; + SysScanDesc lblpropscan; + HeapTuple tup; + Form_pg_propgraph_label_property plpform; + ObjectAddress oa; + + lblpropDesc = table_open(PropgraphLabelPropertyRelationId, + AccessShareLock); + ScanKeyInit(&skey[0], + Anum_pg_propgraph_label_property_oid, + BTEqualStrategyNumber, F_OIDEQ, + ObjectIdGetDatum(object->objectId)); + + lblpropscan = systable_beginscan(lblpropDesc, PropgraphLabelPropertyObjectIndexId, + true, NULL, 1, skey); + + tup = systable_getnext(lblpropscan); + if (!HeapTupleIsValid(tup)) + { + if (!missing_ok) + elog(ERROR, "could not find tuple for label property %u", + object->objectId); + + systable_endscan(lblpropscan); + table_close(lblpropDesc, AccessShareLock); + break; + } + + plpform = (Form_pg_propgraph_label_property) GETSTRUCT(tup); + + ObjectAddressSet(oa, PropgraphElementLabelRelationId, plpform->plpellabelid); + appendStringInfoString(&buffer, getObjectIdentityParts(&oa, objname, + objargs, false)); + + systable_endscan(lblpropscan); + table_close(lblpropDesc, AccessShareLock); + break; + } + case PropgraphLabelRelationId: { HeapTuple tup; diff --git a/src/test/regress/expected/create_property_graph.out b/src/test/regress/expected/create_property_graph.out index bc9a596ec89..eb9e17fb727 100644 --- a/src/test/regress/expected/create_property_graph.out +++ b/src/test/regress/expected/create_property_graph.out @@ -752,6 +752,84 @@ SELECT (pg_identify_object(classid, objid, objsubid)).* type | create_property_graph_tests | g2 | create_property_graph_tests.g2 (21 rows) +-- test object address functions with recursive dependency walk +-- covers all property graph object types including indirect dependents +SELECT * +FROM ( + WITH RECURSIVE deps (classid, objid, objsubid, refclassid, refobjid, refobjsubid) AS ( + SELECT classid, objid, objsubid, + refclassid, refobjid, refobjsubid + FROM pg_depend + WHERE refclassid = 'pg_class'::regclass AND + refobjid = 'create_property_graph_tests.gt'::regclass + UNION ALL + SELECT d.classid, d.objid, d.objsubid, + d.refclassid, d.refobjid, d.refobjsubid + FROM pg_depend d + JOIN deps dp ON d.refclassid = dp.classid AND d.refobjid = dp.objid AND d.refobjsubid = dp.objsubid + ) + SELECT pg_describe_object(classid, objid, objsubid) as obj, + pg_identify_object(classid, objid, objsubid) as ident, + pg_identify_object_as_address(classid, objid, objsubid) as addr + FROM deps +) s +ORDER BY obj COLLATE "C", ident::text COLLATE "C"; + obj | ident | addr +----------------------------------------------------------+----------------------------------------------------------------------------+---------------------------------------------------------------------------- + edge e of property graph gt | ("property graph element",,,"e of create_property_graph_tests.gt") | ("property graph element","{create_property_graph_tests,gt,e}",{}) + edge e of property graph gt | ("property graph element",,,"e of create_property_graph_tests.gt") | ("property graph element","{create_property_graph_tests,gt,e}",{}) + edge e of property graph gt | ("property graph element",,,"e of create_property_graph_tests.gt") | ("property graph element","{create_property_graph_tests,gt,e}",{}) + label e of edge e of property graph gt | ("property graph element label",,,"e of create_property_graph_tests.gt") | ("property graph element label","{create_property_graph_tests,gt,e}",{}) + label e of edge e of property graph gt | ("property graph element label",,,"e of create_property_graph_tests.gt") | ("property graph element label","{create_property_graph_tests,gt,e}",{}) + label e of edge e of property graph gt | ("property graph element label",,,"e of create_property_graph_tests.gt") | ("property graph element label","{create_property_graph_tests,gt,e}",{}) + label e of edge e of property graph gt | ("property graph element label",,,"e of create_property_graph_tests.gt") | ("property graph element label","{create_property_graph_tests,gt,e}",{}) + label e of property graph gt | ("property graph label",,,"e of create_property_graph_tests.gt") | ("property graph label","{create_property_graph_tests,gt,e}",{}) + label v1 of property graph gt | ("property graph label",,,"v1 of create_property_graph_tests.gt") | ("property graph label","{create_property_graph_tests,gt,v1}",{}) + label v1 of vertex v1 of property graph gt | ("property graph element label",,,"v1 of create_property_graph_tests.gt") | ("property graph element label","{create_property_graph_tests,gt,v1}",{}) + label v1 of vertex v1 of property graph gt | ("property graph element label",,,"v1 of create_property_graph_tests.gt") | ("property graph element label","{create_property_graph_tests,gt,v1}",{}) + label v2 of property graph gt | ("property graph label",,,"v2 of create_property_graph_tests.gt") | ("property graph label","{create_property_graph_tests,gt,v2}",{}) + label v2 of vertex v2 of property graph gt | ("property graph element label",,,"v2 of create_property_graph_tests.gt") | ("property graph element label","{create_property_graph_tests,gt,v2}",{}) + label v2 of vertex v2 of property graph gt | ("property graph element label",,,"v2 of create_property_graph_tests.gt") | ("property graph element label","{create_property_graph_tests,gt,v2}",{}) + property a of label v1 of vertex v1 of property graph gt | ("property graph label property",,,"v1 of create_property_graph_tests.gt") | ("property graph label property","{create_property_graph_tests,gt,v1}",{}) + property a of label v1 of vertex v1 of property graph gt | ("property graph label property",,,"v1 of create_property_graph_tests.gt") | ("property graph label property","{create_property_graph_tests,gt,v1}",{}) + property a of label v1 of vertex v1 of property graph gt | ("property graph label property",,,"v1 of create_property_graph_tests.gt") | ("property graph label property","{create_property_graph_tests,gt,v1}",{}) + property a of property graph gt | ("property graph property",,,"a of create_property_graph_tests.gt") | ("property graph property","{create_property_graph_tests,gt,a}",{}) + property b of label v1 of vertex v1 of property graph gt | ("property graph label property",,,"v1 of create_property_graph_tests.gt") | ("property graph label property","{create_property_graph_tests,gt,v1}",{}) + property b of label v1 of vertex v1 of property graph gt | ("property graph label property",,,"v1 of create_property_graph_tests.gt") | ("property graph label property","{create_property_graph_tests,gt,v1}",{}) + property b of label v1 of vertex v1 of property graph gt | ("property graph label property",,,"v1 of create_property_graph_tests.gt") | ("property graph label property","{create_property_graph_tests,gt,v1}",{}) + property b of property graph gt | ("property graph property",,,"b of create_property_graph_tests.gt") | ("property graph property","{create_property_graph_tests,gt,b}",{}) + property c of label e of edge e of property graph gt | ("property graph label property",,,"e of create_property_graph_tests.gt") | ("property graph label property","{create_property_graph_tests,gt,e}",{}) + property c of label e of edge e of property graph gt | ("property graph label property",,,"e of create_property_graph_tests.gt") | ("property graph label property","{create_property_graph_tests,gt,e}",{}) + property c of label e of edge e of property graph gt | ("property graph label property",,,"e of create_property_graph_tests.gt") | ("property graph label property","{create_property_graph_tests,gt,e}",{}) + property c of label e of edge e of property graph gt | ("property graph label property",,,"e of create_property_graph_tests.gt") | ("property graph label property","{create_property_graph_tests,gt,e}",{}) + property c of label e of edge e of property graph gt | ("property graph label property",,,"e of create_property_graph_tests.gt") | ("property graph label property","{create_property_graph_tests,gt,e}",{}) + property c of property graph gt | ("property graph property",,,"c of create_property_graph_tests.gt") | ("property graph property","{create_property_graph_tests,gt,c}",{}) + property k1 of label e of edge e of property graph gt | ("property graph label property",,,"e of create_property_graph_tests.gt") | ("property graph label property","{create_property_graph_tests,gt,e}",{}) + property k1 of label e of edge e of property graph gt | ("property graph label property",,,"e of create_property_graph_tests.gt") | ("property graph label property","{create_property_graph_tests,gt,e}",{}) + property k1 of label e of edge e of property graph gt | ("property graph label property",,,"e of create_property_graph_tests.gt") | ("property graph label property","{create_property_graph_tests,gt,e}",{}) + property k1 of label e of edge e of property graph gt | ("property graph label property",,,"e of create_property_graph_tests.gt") | ("property graph label property","{create_property_graph_tests,gt,e}",{}) + property k1 of label e of edge e of property graph gt | ("property graph label property",,,"e of create_property_graph_tests.gt") | ("property graph label property","{create_property_graph_tests,gt,e}",{}) + property k1 of property graph gt | ("property graph property",,,"k1 of create_property_graph_tests.gt") | ("property graph property","{create_property_graph_tests,gt,k1}",{}) + property k2 of label e of edge e of property graph gt | ("property graph label property",,,"e of create_property_graph_tests.gt") | ("property graph label property","{create_property_graph_tests,gt,e}",{}) + property k2 of label e of edge e of property graph gt | ("property graph label property",,,"e of create_property_graph_tests.gt") | ("property graph label property","{create_property_graph_tests,gt,e}",{}) + property k2 of label e of edge e of property graph gt | ("property graph label property",,,"e of create_property_graph_tests.gt") | ("property graph label property","{create_property_graph_tests,gt,e}",{}) + property k2 of label e of edge e of property graph gt | ("property graph label property",,,"e of create_property_graph_tests.gt") | ("property graph label property","{create_property_graph_tests,gt,e}",{}) + property k2 of label e of edge e of property graph gt | ("property graph label property",,,"e of create_property_graph_tests.gt") | ("property graph label property","{create_property_graph_tests,gt,e}",{}) + property k2 of property graph gt | ("property graph property",,,"k2 of create_property_graph_tests.gt") | ("property graph property","{create_property_graph_tests,gt,k2}",{}) + property m of label v2 of vertex v2 of property graph gt | ("property graph label property",,,"v2 of create_property_graph_tests.gt") | ("property graph label property","{create_property_graph_tests,gt,v2}",{}) + property m of label v2 of vertex v2 of property graph gt | ("property graph label property",,,"v2 of create_property_graph_tests.gt") | ("property graph label property","{create_property_graph_tests,gt,v2}",{}) + property m of label v2 of vertex v2 of property graph gt | ("property graph label property",,,"v2 of create_property_graph_tests.gt") | ("property graph label property","{create_property_graph_tests,gt,v2}",{}) + property m of property graph gt | ("property graph property",,,"m of create_property_graph_tests.gt") | ("property graph property","{create_property_graph_tests,gt,m}",{}) + property n of label v2 of vertex v2 of property graph gt | ("property graph label property",,,"v2 of create_property_graph_tests.gt") | ("property graph label property","{create_property_graph_tests,gt,v2}",{}) + property n of label v2 of vertex v2 of property graph gt | ("property graph label property",,,"v2 of create_property_graph_tests.gt") | ("property graph label property","{create_property_graph_tests,gt,v2}",{}) + property n of label v2 of vertex v2 of property graph gt | ("property graph label property",,,"v2 of create_property_graph_tests.gt") | ("property graph label property","{create_property_graph_tests,gt,v2}",{}) + property n of property graph gt | ("property graph property",,,"n of create_property_graph_tests.gt") | ("property graph property","{create_property_graph_tests,gt,n}",{}) + type gt | (type,create_property_graph_tests,gt,create_property_graph_tests.gt) | (type,{create_property_graph_tests.gt},{}) + type gt[] | (type,create_property_graph_tests,_gt,create_property_graph_tests.gt[]) | (type,{create_property_graph_tests.gt[]},{}) + vertex v1 of property graph gt | ("property graph element",,,"v1 of create_property_graph_tests.gt") | ("property graph element","{create_property_graph_tests,gt,v1}",{}) + vertex v2 of property graph gt | ("property graph element",,,"v2 of create_property_graph_tests.gt") | ("property graph element","{create_property_graph_tests,gt,v2}",{}) +(52 rows) + \a\t SELECT pg_get_propgraphdef('g2'::regclass); CREATE PROPERTY GRAPH create_property_graph_tests.g2 diff --git a/src/test/regress/expected/event_trigger.out b/src/test/regress/expected/event_trigger.out index 065f586310f..ed8d5df397e 100644 --- a/src/test/regress/expected/event_trigger.out +++ b/src/test/regress/expected/event_trigger.out @@ -173,6 +173,23 @@ NOTICE: test_event_trigger: ddl_command_end CREATE USER MAPPING alter default privileges for role regress_evt_user revoke delete on tables from regress_evt_user; NOTICE: test_event_trigger: ddl_command_end ALTER DEFAULT PRIVILEGES +-- DROP PROPERTY GRAPH should work with event trigger in place +CREATE TABLE t1x (a int PRIMARY KEY, b text); +NOTICE: test_event_trigger: ddl_command_start CREATE TABLE +NOTICE: test_event_trigger: ddl_command_end CREATE TABLE +CREATE TABLE t2x (i int PRIMARY KEY, j text); +NOTICE: test_event_trigger: ddl_command_start CREATE TABLE +NOTICE: test_event_trigger: ddl_command_end CREATE TABLE +CREATE PROPERTY GRAPH gx + VERTEX TABLES ( + t1x KEY (a) LABEL l1 PROPERTIES (b AS p1), + t2x KEY (i) LABEL l2 PROPERTIES (j AS p1) +); +NOTICE: test_event_trigger: ddl_command_end CREATE PROPERTY GRAPH +DROP PROPERTY GRAPH gx; +NOTICE: test_event_trigger: ddl_command_end DROP PROPERTY GRAPH +DROP TABLE t1x, t2x; +NOTICE: test_event_trigger: ddl_command_end DROP TABLE -- alter owner to non-superuser should fail alter event trigger regress_event_trigger owner to regress_evt_user; ERROR: permission denied to change owner of event trigger "regress_event_trigger" diff --git a/src/test/regress/expected/object_address.out b/src/test/regress/expected/object_address.out index 97227d67a54..5e53f0b092e 100644 --- a/src/test/regress/expected/object_address.out +++ b/src/test/regress/expected/object_address.out @@ -67,7 +67,9 @@ DECLARE BEGIN FOR objtype IN VALUES ('toast table'), ('index column'), ('sequence column'), ('toast table column'), ('view column'), ('materialized view column'), - ('property graph element'), ('property graph label'), ('property graph property') + ('property graph element'), ('property graph element label'), + ('property graph label'), ('property graph label property'), + ('property graph property') LOOP BEGIN PERFORM pg_get_object_address(objtype, '{one}', '{}'); @@ -84,7 +86,9 @@ WARNING: error for toast table column: unsupported object type "toast table col WARNING: error for view column: unsupported object type "view column" WARNING: error for materialized view column: unsupported object type "materialized view column" WARNING: error for property graph element: unsupported object type "property graph element" +WARNING: error for property graph element label: unrecognized object type "property graph element label" WARNING: error for property graph label: unsupported object type "property graph label" +WARNING: error for property graph label property: unrecognized object type "property graph label property" WARNING: error for property graph property: unsupported object type "property graph property" -- miscellaneous other errors select * from pg_get_object_address('operator of access method', '{btree,integer_ops,1}', '{int4,bool}'); @@ -593,7 +597,9 @@ WITH objects (classid, objid, objsubid) AS (VALUES ('pg_parameter_acl'::regclass, 0, 0), -- no parameter ACL ('pg_policy'::regclass, 0, 0), -- no policy ('pg_propgraph_element'::regclass, 0, 0), -- no property graph element + ('pg_propgraph_element_label'::regclass, 0, 0), -- no property graph element label ('pg_propgraph_label'::regclass, 0, 0), -- no property graph label + ('pg_propgraph_label_property'::regclass, 0, 0), -- no property graph label property ('pg_propgraph_property'::regclass, 0, 0), -- no property graph property ('pg_publication'::regclass, 0, 0), -- no publication ('pg_publication_namespace'::regclass, 0, 0), -- no publication namespace @@ -653,6 +659,8 @@ ORDER BY objects.classid, objects.objid, objects.objsubid; ("(""parameter ACL"",,,)")|("(""parameter ACL"",,)")|NULL ("(""property graph element"",,,)")|("(""property graph element"",,)")|NULL ("(""property graph label"",,,)")|("(""property graph label"",,)")|NULL +("(""property graph element label"",,,)")|("(""property graph element label"",,)")|NULL ("(""property graph property"",,,)")|("(""property graph property"",,)")|NULL +("(""property graph label property"",,,)")|("(""property graph label property"",,)")|NULL -- restore normal output mode \a\t diff --git a/src/test/regress/sql/create_property_graph.sql b/src/test/regress/sql/create_property_graph.sql index 241f93df302..9608dd5c4af 100644 --- a/src/test/regress/sql/create_property_graph.sql +++ b/src/test/regress/sql/create_property_graph.sql @@ -298,6 +298,29 @@ SELECT (pg_identify_object(classid, objid, objsubid)).* refobjid = 'create_property_graph_tests.g2'::regclass ORDER BY 1, 2, 3, 4; +-- test object address functions with recursive dependency walk +-- covers all property graph object types including indirect dependents +SELECT * +FROM ( + WITH RECURSIVE deps (classid, objid, objsubid, refclassid, refobjid, refobjsubid) AS ( + SELECT classid, objid, objsubid, + refclassid, refobjid, refobjsubid + FROM pg_depend + WHERE refclassid = 'pg_class'::regclass AND + refobjid = 'create_property_graph_tests.gt'::regclass + UNION ALL + SELECT d.classid, d.objid, d.objsubid, + d.refclassid, d.refobjid, d.refobjsubid + FROM pg_depend d + JOIN deps dp ON d.refclassid = dp.classid AND d.refobjid = dp.objid AND d.refobjsubid = dp.objsubid + ) + SELECT pg_describe_object(classid, objid, objsubid) as obj, + pg_identify_object(classid, objid, objsubid) as ident, + pg_identify_object_as_address(classid, objid, objsubid) as addr + FROM deps +) s +ORDER BY obj COLLATE "C", ident::text COLLATE "C"; + \a\t SELECT pg_get_propgraphdef('g2'::regclass); SELECT pg_get_propgraphdef('g3'::regclass); diff --git a/src/test/regress/sql/event_trigger.sql b/src/test/regress/sql/event_trigger.sql index 32e9bb58c5e..271f443378b 100644 --- a/src/test/regress/sql/event_trigger.sql +++ b/src/test/regress/sql/event_trigger.sql @@ -143,6 +143,19 @@ create user mapping for regress_evt_user server useless_server; alter default privileges for role regress_evt_user revoke delete on tables from regress_evt_user; +-- DROP PROPERTY GRAPH should work with event trigger in place +CREATE TABLE t1x (a int PRIMARY KEY, b text); +CREATE TABLE t2x (i int PRIMARY KEY, j text); + +CREATE PROPERTY GRAPH gx + VERTEX TABLES ( + t1x KEY (a) LABEL l1 PROPERTIES (b AS p1), + t2x KEY (i) LABEL l2 PROPERTIES (j AS p1) +); + +DROP PROPERTY GRAPH gx; +DROP TABLE t1x, t2x; + -- alter owner to non-superuser should fail alter event trigger regress_event_trigger owner to regress_evt_user; diff --git a/src/test/regress/sql/object_address.sql b/src/test/regress/sql/object_address.sql index 1bbe9457c1c..27109d421c7 100644 --- a/src/test/regress/sql/object_address.sql +++ b/src/test/regress/sql/object_address.sql @@ -67,7 +67,9 @@ DECLARE BEGIN FOR objtype IN VALUES ('toast table'), ('index column'), ('sequence column'), ('toast table column'), ('view column'), ('materialized view column'), - ('property graph element'), ('property graph label'), ('property graph property') + ('property graph element'), ('property graph element label'), + ('property graph label'), ('property graph label property'), + ('property graph property') LOOP BEGIN PERFORM pg_get_object_address(objtype, '{one}', '{}'); @@ -282,7 +284,9 @@ WITH objects (classid, objid, objsubid) AS (VALUES ('pg_parameter_acl'::regclass, 0, 0), -- no parameter ACL ('pg_policy'::regclass, 0, 0), -- no policy ('pg_propgraph_element'::regclass, 0, 0), -- no property graph element + ('pg_propgraph_element_label'::regclass, 0, 0), -- no property graph element label ('pg_propgraph_label'::regclass, 0, 0), -- no property graph label + ('pg_propgraph_label_property'::regclass, 0, 0), -- no property graph label property ('pg_propgraph_property'::regclass, 0, 0), -- no property graph property ('pg_publication'::regclass, 0, 0), -- no publication ('pg_publication_namespace'::regclass, 0, 0), -- no publication namespace -- 2.34.1 --R84XrR4Hgw0dH3Ec-- ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-06-16 11:54 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-06-16 11:54 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2268cc277bc..127d4415084 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -609,11 +608,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -627,11 +623,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -705,6 +696,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -881,11 +874,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -901,8 +892,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -922,6 +912,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -932,33 +924,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2346,21 +2341,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2372,6 +2391,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2380,76 +2402,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v01-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
* [PATCH 1/8] Use tuple slot to pass tuples for rewriting. @ 2026-07-15 08:37 Antonin Houska <ah@cybertec.at> 0 siblings, 0 replies; 604+ messages in thread From: Antonin Houska @ 2026-07-15 08:37 UTC (permalink / raw) This patch tries to adopt the preferable way of handling tuples, i.e. pass the containing tuple slot rather than the actual tuple. The motivation is that heap_insert_for_repack() will need a slot in the near future, in order to call ExecInsertIndexTuples(). Also, in order to set values of dropped attributes to to NULL, it seems more compact to pass a tuple slot as a workspace than two arrays (one for values and one for nulls). --- src/backend/access/heap/heapam_handler.c | 199 +++++++++++++---------- 1 file changed, 110 insertions(+), 89 deletions(-) diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index bf87430cf01..4db8a16f691 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -48,14 +48,13 @@ #include "utils/rel.h" #include "utils/tuplesort.h" -static void reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate); -static void heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull, - BulkInsertState bistate); -static HeapTuple reform_tuple(HeapTuple tuple, Relation OldHeap, - Relation NewHeap, Datum *values, bool *isnull); +static void reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate); +static void heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, + BulkInsertStateData *bistate); +static bool tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc); +static void clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform); static bool SampleHeapTupleVisible(TableScanDesc scan, Buffer buffer, HeapTuple tuple, @@ -603,11 +602,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, bool is_system_catalog; Tuplesortstate *tuplesort; TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); TupleTableSlot *slot; - int natts; - Datum *values; - bool *isnull; + TupleTableSlot *reform_slot; BufferHeapTupleTableSlot *hslot; BlockNumber prev_cblock = InvalidBlockNumber; bool concurrent = snapshot != NULL; @@ -621,11 +617,6 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, */ Assert(RelationGetTargetBlock(NewHeap) == InvalidBlockNumber); - /* Preallocate values/isnull arrays */ - natts = newTupDesc->natts; - values = palloc_array(Datum, natts); - isnull = palloc_array(bool, natts); - /* * In non-concurrent mode, initialize the rewrite operation. This is not * needed in concurrent mode. @@ -699,6 +690,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, slot = table_slot_create(OldHeap, NULL); hslot = (BufferHeapTupleTableSlot *) slot; + reform_slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsVirtual); /* * Scan through the OldHeap, either in OldIndex order or sequentially; @@ -875,11 +868,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, int64 ct_val[2]; if (!concurrent) - reform_and_rewrite_tuple(tuple, OldHeap, NewHeap, - values, isnull, rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* * In indexscan mode and also VACUUM FULL, report increase in @@ -895,8 +886,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, index_endscan(indexScan); if (tableScan != NULL) table_endscan(tableScan); - if (slot) - ExecDropSingleTupleTableSlot(slot); + ExecDropSingleTupleTableSlot(slot); /* * In scan-and-sort mode, complete the sort, then read out all live tuples @@ -916,6 +906,8 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, pgstat_progress_update_param(PROGRESS_REPACK_PHASE, PROGRESS_REPACK_PHASE_WRITE_NEW_HEAP); + slot = MakeSingleTupleTableSlot(RelationGetDescr(OldHeap), + &TTSOpsHeapTuple); for (;;) { HeapTuple tuple; @@ -926,33 +918,36 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap, if (tuple == NULL) break; + /* + * XXX Ideally we should use tuplesort_gettupleslot() above, but + * it retrieves minimal tuples and tuplesort_puttupleslot() cannot + * get the tuple descriptor from tuplesort created by + * tuplesort_begin_cluster(). + */ + ExecStoreHeapTuple(tuple, slot, false); + n_tuples += 1; if (!concurrent) - reform_and_rewrite_tuple(tuple, - OldHeap, NewHeap, - values, isnull, - rwstate); + reform_and_rewrite_tuple(slot, reform_slot, rwstate); else - heap_insert_for_repack(tuple, OldHeap, NewHeap, - values, isnull, bistate); + heap_insert_for_repack(NewHeap, slot, reform_slot, bistate); /* Report n_tuples */ pgstat_progress_update_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, n_tuples); } + ExecDropSingleTupleTableSlot(slot); tuplesort_end(tuplesort); } + ExecDropSingleTupleTableSlot(reform_slot); + /* Write out any remaining tuples, and fsync if needed */ if (rwstate) end_heap_rewrite(rwstate); if (bistate) FreeBulkInsertState(bistate); - - /* Clean up */ - pfree(values); - pfree(isnull); } /* @@ -2340,21 +2335,45 @@ heapam_scan_sample_next_tuple(TableScanDesc scan, SampleScanState *scanstate, * currently only known to happen as an after-effect of ALTER TABLE * SET WITHOUT OIDS. * - * So, we must reconstruct the tuple from component Datums. + * So, we must reconstruct the tuple from component Datums. 'reform' slot + * is a workspace for this reconstruction. */ static void -reform_and_rewrite_tuple(HeapTuple tuple, - Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, RewriteState rwstate) +reform_and_rewrite_tuple(TupleTableSlot *src, TupleTableSlot *reform, + RewriteState rwstate) { - HeapTuple newtuple; + HeapTuple tuple, + newtuple; + bool shouldFree, + shouldFreeNew; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + /* + * The old tuple will not be modified, so do not request materialization. + * (A copy can be created for specific slot type though, e.g. + * TTSOpsMinimalTuple.) + */ + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + + /* No need to materialize, copy will be created anyway. */ + Assert(TTS_IS_VIRTUAL(reform)); + newtuple = ExecFetchSlotHeapTuple(reform, false, &shouldFreeNew); + } + else + { + newtuple = heap_copytuple(tuple); + shouldFreeNew = true; + } /* The heap rewrite module does the rest */ rewrite_heap_tuple(rwstate, tuple, newtuple); - heap_freetuple(newtuple); + if (shouldFree) + heap_freetuple(tuple); + if (shouldFreeNew) + heap_freetuple(newtuple); } /* @@ -2366,6 +2385,9 @@ reform_and_rewrite_tuple(HeapTuple tuple, * information). Thus we must use heap_insert() both during the * catch-up and here. * + * 'reform' is a slot to use for tuple "reforming", typically to get set + * values of dropped columns to NULL. + * * We pass the NO_LOGICAL flag to heap_insert() in order to skip logical * decoding: as soon as REPACK CONCURRENTLY swaps the relation files, it drops * this relation, so no logical replication subscription should need the data. @@ -2374,76 +2396,75 @@ reform_and_rewrite_tuple(HeapTuple tuple, * case. */ static void -heap_insert_for_repack(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull, BulkInsertState bistate) +heap_insert_for_repack(Relation rel, TupleTableSlot *src, + TupleTableSlot *reform, BulkInsertStateData *bistate) { - HeapTuple newtuple; + HeapTuple tuple; + bool shouldFree; + TupleTableSlot *slot; - newtuple = reform_tuple(tuple, OldHeap, NewHeap, values, isnull); + tuple = ExecFetchSlotHeapTuple(src, false, &shouldFree); + if (tuple_needs_reform(tuple, src->tts_tupleDescriptor)) + { + clear_dropped_attributes(tuple, reform); + slot = reform; + } + else + slot = src; - heap_insert(NewHeap, newtuple, GetCurrentCommandId(true), - HEAP_INSERT_NO_LOGICAL, bistate); + /* + * clear_dropped_attributes() should have deformed the tuple, so nothing + * should depend on it now. + */ + if (shouldFree) + heap_freetuple(tuple); - heap_freetuple(newtuple); + table_tuple_insert(rel, slot, GetCurrentCommandId(true), + TABLE_INSERT_NO_LOGICAL, bistate); } -/* - * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. - * - * Deform the given tuple, set values of dropped columns to NULL, and fill in - * any values from attmissingval; then form a new tuple and return it. If no - * attributes need to be changed, a copy of the original tuple is returned. - * Caller is responsible for freeing the returned tuple. - * - * XXX this coding assumes that both relations have the same tupledesc. - */ -static HeapTuple -reform_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, - Datum *values, bool *isnull) +static bool +tuple_needs_reform(HeapTuple tuple, TupleDesc tupDesc) { - TupleDesc oldTupDesc = RelationGetDescr(OldHeap); - TupleDesc newTupDesc = RelationGetDescr(NewHeap); - bool needs_reform = false; - /* * A short tuple might require values from attmissing val, so activate the * coding unconditionally in that case. The value might legitimally be * NULL otherwise, so this is slightly wasteful, but it probably beats * having to test each attribute for presence of attmissingval each time. */ - if (HeapTupleHeaderGetNatts(tuple->t_data) < newTupDesc->natts) - needs_reform = true; + if (HeapTupleHeaderGetNatts(tuple->t_data) < tupDesc->natts) + return true; - /* - * If the column has been dropped but a value is still present, we can - * optimize storage now by getting rid of it. - */ - if (!needs_reform) + /* Does it have dropped attributes? */ + for (int i = 0; i < tupDesc->natts; i++) { - for (int i = 0; i < newTupDesc->natts; i++) - { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped && - !heap_attisnull(tuple, i + 1, newTupDesc)) - { - needs_reform = true; - break; - } - } + if (TupleDescCompactAttr(tupDesc, i)->attisdropped && + !heap_attisnull(tuple, i + 1, tupDesc)) + return true; } - /* Skip work if no changes are needed */ - if (!needs_reform) - return heap_copytuple(tuple); + return false; +} - heap_deform_tuple(tuple, oldTupDesc, values, isnull); +/* + * Subroutine for reform_and_rewrite_tuple and heap_insert_for_repack. + * + * Set values of dropped columns to NULL, + */ +static void +clear_dropped_attributes(HeapTuple tuple, TupleTableSlot *reform) +{ + TupleDesc tupDesc = reform->tts_tupleDescriptor; + + /* Assuming 'reform' is virtual, this deforms the tuple. */ + Assert(TTS_IS_VIRTUAL(reform)); + ExecForceStoreHeapTuple(tuple, reform, false); - for (int i = 0; i < newTupDesc->natts; i++) + for (int i = 0; i < tupDesc->natts; i++) { - if (TupleDescCompactAttr(newTupDesc, i)->attisdropped) - isnull[i] = true; + if (TupleDescCompactAttr(tupDesc, i)->attisdropped) + reform->tts_isnull[i] = true; } - - return heap_form_tuple(newTupDesc, values, isnull); } /* -- 2.52.0 --=-=-= Content-Type: text/x-diff Content-Disposition: attachment; filename=v02-0002-Move-functions-to-repack.c.patch ^ permalink raw reply [nested|flat] 604+ messages in thread
end of thread, other threads:[~2026-07-15 08:37 UTC | newest] Thread overview: 604+ messages (download: mbox mbox.gz follow: Atom feed) -- links below jump to the message on this page -- 2026-04-23 09:27 [PATCH v3] Fix DROP PROPERTY GRAPH "unsupported object class" error Bertrand Drouvot <bertranddrouvot.pg@gmail.com> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-06-16 11:54 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at> 2026-07-15 08:37 [PATCH 1/8] Use tuple slot to pass tuples for rewriting. Antonin Houska <ah@cybertec.at>
This inbox is served by agora; see mirroring instructions for how to clone and mirror all data and code used for this inbox